Extract Photo & Signature from Scanned PDF Files (Python / Image Processing)

I need a reliable script/system that can extract photo and signature from scanned PDF files.
These PDFs are fully scanned documents (no text layer) where photo and signature are part of the image.

Key Requirements (Read Carefully):

PDFs are scanned (image-based), not digital

Layout is mostly the same, but photo/signature boxes may shift slightly

Photo & signature are usually on page 1

If box detection fails, fallback to face detection for photo is required

Signature extraction should work using image processing (ink/contour based)

Output files must use same base filename as the PDF

Example: ABC123.pdf

ABC123_photo.jpg

ABC123_sign.png

Technical Expectations:

Language: Python (preferred)

Libraries: OpenCV, PyMuPDF / pdf2image, NumPy (or equivalent)

DPI handling (300–400 DPI)

Deskew / preprocessing for scanned PDFs

Batch processing (folder-based)

Clean, reusable, well-structured code

Nice to Have (Bonus):

Config-driven ROI (JSON instead of hardcoded values)

Logging (success / failure)

Debug images for failed cases

Ability to handle multiple pages if page 1 fails

What I Will Provide:

Sample scanned PDFs

Clear expected output examples

Feedback during development

Who Should Apply:

You have real experience with image processing

You understand scanned documents (not OCR-only solutions)

You can deliver working logic, not just demo scripts

Do NOT apply if:

You only work with text-based PDFs

You rely only on OCR

You have no OpenCV/image-processing experience

Budget:
Open to reasonable bids based on solution quality.